Michael Heldmann, Ph.D., CFA
Managing Director, Sr. Portfolio Manager, Best Styles US Equity Head of Best Styles N.A.
Allianz Global Investors
Investors are adding a new criterion to the list of factors they must assess when reviewing a new investment strategy or manager: Does the strategy use artificial intelligence (A.I.), and if so, does the technology add value? Answering the second of these two questions is not easy, but it helps to start with one simple premise: A.I. is a tool, not a destination.
You can learn more about the Forum here.
- It’s important to remember that A.I. is really nothing but pattern recognition
- The traditional use of data for asset management collects a large set of structured data (i.e. rate of return, market value, etc.) in an attempt to understand your investments and predict future performance.
- The development of A.I. has allowed asset managers to use large amounts of unstructured data (i.e. pictures, graphs, social media, etc.) and convert that information into structured, qualitative data.
- Using machine learning and A.I., the unstructured data can be integrated with traditional data sets for excess return assumptions, risk modeling, etc.
- The growing awareness of A.I. within asset management, and the rest of the word, has been due to the availability of necessary computing power. A.I. needs a certain level of computing power to process and learn from data it collects. We are reaching that point.
- For A.I. to be effective, it needs a tremendous amount of data. Systems need billions upon billions of data points to process and learn.
- Much more data will be needed before A.I. can be used effectively for forecasting. For example, forecasting a three-month return: we currently only have 1.2 million observations to train the system. Well short of the several billions needed.
- While A.I. may never replace the full Asset Management process, including portfolio management decisions and research, there will always be a role for A.I. within the investment process.
- Currently, an area where A.I. has benefited investors is the processing of earnings calls. Each quarter, thousands of earnings calls take place, which have useful information, but cannot easily be integrated within data analysis. A.I. can automate the process of analyzing these calls and turning them into structured data.
- It will be key to integrate A.I. whenever it is beneficial in the investment process, but to also keep many existing structures; where we have great understanding of how they work and what they mean really for the portfolio.